The first computation is the dip-steering attribute

The first computation is the dip-steering attribute, an essential product for the calculation of most of the seismic attributes in OpendTect. It shows the parameter tests including the Fourier Transform algorithm and the BG algorithm The BG algorithm shows of the quality of the Fourier Transform algorithm, for a fraction of the computational time. The low seismic resolution and high sample rate justifies the tradeoff of losing of accuracy in exchange for speed. A further approach could include resampling the seismic data toprior the attribute extraction.
The main disadvantage is that BG algorithm needs a median filter to smooth the results. To compensate this, shows the extraction workflow that includes two volumes. Dip-steering testing parameters summarize the testing parameters. The first volume is a detailed dip-steering, which focuses on the high-frequency events and features a median filter. The second volume filters even more in the three directions to keep the structural trend of the events. summarizes the best combination.
The first seismic attribute is Similarity; shows the test parameter. The dip-steering shows better performance over None and Browse dip. An smaller time gate highlights high-frequency events, while bigger time gate smooths the results, suitable for low-frequency trends
shows that section is less noisy than. The average output ranks better than the minimum, while the maximum statistic output turns zero for no dip-steering cubes. In sum, Error! Reference source not found. shows the best results, for low and high resolution, and the final output uses the SEGY format to export and import into the HRS platform.
The second attribute is Curvature, the most positive and most negative curvatures are the subject of investigation, whereas other curvature families are not taking part in the analysis, after the suggestion from literature towards tight gas reservoir characterization. This attribute includes the analysis with both dip-steering algorithms, with a constant velocity of. summarizes the testing parameters. The results show that the step out of 2 has a better approximation for high-frequency events, and a step out of 7, handles the low-frequency components better.
shows the comparison between the dip-steering algorithms. FFT Precise shows better results, nevertheless is inefficient for its extended computational time. Therefore, the BG algorithm gives convenience, reducing the time for acceptable quality. On the whole, sum up the best results. In future works, the FFT Precise algorithm can add accuracy to the study. In the end, OpendTect generates the SEGY volumes for the multi-attribute analysis in HRS.
The third attribute is the GLCM textures, this analysis is more comprehensive and includes a higher number of variables compared with previous attributes. The GLCM textures in this study are Contrast, Entropy, Dissimilarity, and Homogeneity, because of its advantages for tight gas sandstone reservoirs. The other textures have spurious relationships of independence. Therefore, the present work excludes further investigation in that direction.
The GLCM parameter tests are summed up in ; it also includes the amplitude range around the target area. The GLCM directional tests spend considerable time because there are 14 directions, according to. Single direction detail analysis gives detailed information about fracturing and facies distribution, while all directions is a smoother and miss the subtle features. As a result, each direction have 24 attributes, as shown in with gray levels vary from 16- to 256-bit and dip-steering volumes. Therefore, the total combinations are 4680. For this reason, the study limit to the full directions. Nevertheless, a more exhaustive texture-oriented analysis, according to the direction of the channels open the window for future works.
compares the GLCM attributes at ms. It is clear how attributes uncover the channels hidden in the original data. shows the Entropy analysis for different directions at 1880 ms. Additionally, the GLCM Mean shows good independent results. compares the GLCM Mean with the GLCM Cluster shade with 16 gray levels, one trace for the analysis in each direction (inline, crossline, and time), and a vertical search window o.
In the end, the best attributes have a step out of 1 trace in the inline, crossline and time direction for a higher resolution, because higher step out refers to low frequencies. Also, 5-bit gray levels, a matrix of 32×32 to increase the resolution. Even though the results were indistinct to changes in the time gate size, the default window is ±8ms. The dip-steering takes greater computational time versus the conventional approach, assuredly the results shows neither variations nor significant improvements. Consequently, the optimization does not consider it. In the end, OpendTect extracts the volumes in SEGY, as previous attribute steps, for HRS.
Furthermore, the study includes corresponding attributes related to the fluid characteristic. The Weighted Envelope Frequency shows longer wavelength variations which relates to gas bodies. The Thin Bed Indicator is the difference between instantaneous frequency and Envelope weighted frequency. Moreover, the Energy, the squared sum of the samples values in the specific time-gate divided by the number of samples in the gate. There were 4-time windows in ms that offers good independent responses, including gradient computation.
The first step is to predict P-wave and Density attributes, to certify the reliability of the workflow. The second step is the permeability prediction. summarizes the external attributes for the reservoir property predictions; it includes 22 volumes (22 attributes extracted + Acoustic Impedance + Raw Seismic), plus the 18 internal attributes of
shows the response of the 22 volumes at the well . includes the base parameters for the reservoir properties prediction in the HRS EMERGE platform. The first range focus on the reservoir area and the second covers an extended zone to apply extended operators successfully.
Moreover, the well correlation is an essential factor for the prediction. Therefore, the method works with three datasets, the original with eight wells; the set number 2 with ten wells with more than 60% of correlation; and similarly set number 3, with eight wells of correlations above. Other factors include the validation of search criteria, the removal of following attributes by error improvement, the pre-whitening factor set at and the operator lag fix in 0.